Tropical Cyclone Wind Characteristics for the Bangladesh Coast Using Monte Carlo Simulation

P 4.12 Tropical Cyclone Wind Characteristics for the Bangladesh Coast Using Monte Carlo Simulation

Tanveerul Islam and Richard E. Peterson* Wind Science and Engineering Research Center, Texas Tech University, Lubbock, Texas.

Introduction The Sites Wind and storm surge are crucial factors The sites chosen for the simulations are: in the determination of how much damage occurs Khulna, Barisal, Noakhali, Chittagong and Cox’s in the coastal regions in association with any Bazar. (Figure-1) particular hurricane or tropical cyclone. Since very few tropical storms actually strike at a particular site and historical records only exist for the past hundred years or so, there exist a limited number of observations at each site. This is a major problem when considering cyclone wind statistics for research. Batts et al. (1980) determined that it is not possible to use statistical analysis of the highest annual wind speeds at a particular site in order to determine the extreme wind speeds in hurricane- prone regions. Monte Carlo simulation is, therefore, used in this study as an indirect method to obtain the cyclone wind statistics for the Bangladesh coast.

The Holland (1991) wind field model is chosen and used to complete the Monte Carlo simulations. The essential parameters needed to Figure-1 Sites used in simulations generate a hurricane wind field in the Holland model are: storm heading or angle of attack, forward translation speed, radius of the Methodology maximum wind and central pressure. This paper In order to use the primary data in the describes the methodology of the Monte Carlo simulation probability density functions of the simulation and the probability distributions of storm parameters must be specified. Best fit 4.5 the hurricane wind field parameters used in this software from Palisade is used to find the best fit study. distributions of the primary data. After obtaining the best fit distribution, a probability distribution Data Resources table for each parameter is constructed (Table 1- The primary data collected for central Table 4). pressure dated from 1971 to 2003 and for the radius of maximum wind dated from 1981 to The programs for the Monte Carlo 2000. The corresponding data for the earlier simulation are written in FORTRAN. The first storms are not available. The data of the radius program randomly selects the four parameters of maximum wind are obtained from the needed to generate a storm and creates an output Statistical Year Book of Bangladesh 2000 file of the 1000 randomly generated variables. provided by The Bangladesh Meteorological Department (BMD). Data for other parameters The second program runs the actual are obtained from the Global Tropical Cyclone Monte Carlo simulation for each site using the Climatic Atlas (GTCCA 1.0) database. The storm information generated by the first program primary data collected for angle of attack and and the site information, which includes the forward translation speed are dated back from bathymetry of the Bay of Bengal obtained from the 1877 to 2003. nautical chart. This program generates 1000 simulated storms with maximum wind speed and ______*Corresponding author address: Richard E. Peterson, Texas Tech University, Dept. of Geosciences, Lubbock, TX 79409-2101. em ail: [email protected] 1 surge height at landfall. Each storm is initially of selecting a speed of 10.08, a 20% chance of generated offshore a distance, which requires 20 selecting a speed of 7.06, 4.88 and 2.89 knots, a hours for the storm to move ashore. After the 15% chance of selecting a speed of 1.14 knots, landfall, the simulation lasts for 10 more hours in and a 5% chance that a forward speed of 1.0 inland. knots is selected for any simulation. This same set of steps is used in selecting the forward speed Data Distributions for all other sites. Three different methods are used to find the best fit distributions. These are: Chi-Square, Since each site individually has too few Anderson-Darling and K-S methods. data for the radius of the maximum wind (RMW) and central pressure, it is difficult to get any Table-1 Storm heading distributions. Units are distribution from the data for each site. In order in degrees clockwise from north. The to get a distribution, data for the sites located at percentages represent the probability the storm the western part (Khulna, Barisal and Noakhali) heading will be less than the given value. are combined together under ‘west coast’ and for the eastern part (Chittagong and Cox’s Bazar) under ‘east coast.’ These distributions are then Lower Site 95% 83.33% 50% 16.67% 5% bound used for each site according to their respective Khulna 359 345 186 19.3 3.9 3 location. Barisal 359 354 183 21.6 3 Noakhali 203.6 57.5 19.94 11.18 7 Table-3 RMW distributions. Units are in Chittagong 151 97 60 40 30.76 17 nautical mile. Cox'sBazar 86 68.45 43.27 18.09 9 Lower Site 95% 83.33% 50% 16.67% 5% bound West From Table-1, at Khulna the storm Coast 42.78 39.39 34.56 29.72 26.34 10 heading is sampled so that there is a 5% chance East Coast 46.08 41.98 37.03 32.08 27.98 10 of selecting a storm heading of 359°, a 11.67% chance of selecting a storm heading of 345°, a 33.33% chance of selecting a storm heading of From Table-3, at ‘west coast’, i.e. for 186° and 19.3°, a 11.67% chance of selecting a Khulna, Barisal and Noakhali, the sample was storm heading of 3.9°, and a 5% chance that a such that there is a 5% chance of selecting storm heading of 3° will be selected for any radius of maximum wind of 42.78 nmi, a 11.67% simulation (the directions are in the chance of selecting a radius of maximum wind of meteorological system, where 0° was a storm 39.39 nmi, a 33.33% chance of selecting a radius which moved out of the true north, 90° was a of maximum wind of 34.56 nmi and 29.72 nmi, a storm which moved from east to west, 180° was 11.67% chance of selecting a radius of maximum a storm which moved from south to north, etc.). wind of 26.34 nmi and a 5% chance that a radius This same process is used in selecting the storm of maximum wind of 10.0 nmi for any heading for all of the other sites. simulation. The same process is used in selecting the radius of maximum winds for the sites in Table-2 Forward speed distributions of five ‘east coast’. locations. Units are in knots. Table-4 Central pressure distributions. Units are Site 95% 80% 60% 40% 20% 5% L.B. in milibars.

Khulna 14.93 10.08 7.06 4.88 2.89 1.14 1 Site 90% 70% 50% 30% 15% 5% West Barisal 15.19 12.09 9.82 7.87 5.59 2.49 2 Coast 1001.22 987.56 978.09 968.82 959.38 948 Noakhali 15.48 12.55 10.39 8.55 6.39 3.46 2 East Coast 1014.56 985.72 965.75 945.78 926.27 903 Chittgong 24.47 18.25 15.35 13.41 11.51 9.27 4.2 Cox's Bazar 19.85 15.66 12.61 10.08 7.34 4.17 2 From Table-4, the central pressure is sampled for the sites through the process as From Table-2, at Khulna the forward described above. speed is sampled so that there is a 5% chance of selecting a speed of 14.93 knots, a 15% chance References: Available upon request.